A/B Testing an email campaign and getting experiment results on downstream product metrics (in addition to top level email interaction metrics), is a common use case for Statsig customers.
The general flow looks like this:
- Create an Experiment in Statsig for this email campaign.
- Check that experiment for each user you are sending the email to using a server SDK (very performant for large volumes). Record the variant of the email to send. If you are writing a script to do this, make sure to send the exposure events to Statsig (either
shutdown()the SDK before exiting the script). If you are unsure, check the Diagnostics page of your experiment and make sure the exposure checks are recorded for each of the users.
- Using the list you have of users for each variant, send the corresponding email
- Downstream events you log on your client or server application must use the same ID and ID type as the experiment check in step (2).
#4 is the biggest gotcha. When you check the experiment for each of the users you need to pass a user object to Statsig. If your experiment is on the userID id type, you need to make sure each of those checks sets the userID on the user object. Any metrics and events you have instrumented in your app/website/server must pass that same userID in order for Statsig to join product metrics with experiment variant assignment.